Creating robust evaluation harnesses ensures that AI models are tested against realistic software engineering scenarios. This work directly feeds into improving how autonomous agents handle complex coding objectives.

During this remote session, you will review a series of coding tasks and their corresponding evaluation harnesses. You will assess whether the programming challenges accurately reflect real-world software engineering problems. We will ask you to verify the logic, test cases, and overall structure of the environments provided. Your technical feedback will be captured through a guided conversation and screen-sharing exercises.

• You will be engaged as an independent contractor.
• Opportunities can be extended, shortened, or concluded early depending on needs and performance.
• Your participation will not involve access to confidential or proprietary information from any employer, client, or institution.

• Review realistic programming tasks for accuracy and difficulty.
• Verify the logic and test cases within provided evaluation harnesses.
• Assess whether coding environments effectively measure software engineering skills.
• Walk us through your thought process while analyzing complex code structures.

We are looking for practicing software engineers with strong backgrounds in building and testing complex systems. Candidates should have direct experience writing test suites, evaluation harnesses, or comprehensive code reviews. We welcome backend engineers, full-stack developers, machine learning engineers, and software architects.

• Active professional experience as a software engineer.
• Hands-on experience building or maintaining test suites and evaluation harnesses.
• Comfortable reviewing code and explaining technical concepts aloud.
• Familiarity with complex real-world software architecture.
• Currently works as a Software Engineer (backend, full-stack, frontend), ML or AI Engineer, or authors coding tasks for AI evaluation.
• Has at least 3 years of paid, full-time experience writing production software.
• Has personally authored coding tasks with issue descriptions and test suites for AI training or evaluation.
• Must have experience with benchmarks, harnesses, or sandboxing frameworks used to train or evaluate software-engineering agents.

They will answer questions about their experience building AI coding tasks and complete an assessment. This takes about 15 minutes.

You'll start with a short screening interview.

Work arrangement: Fully remote

Eligible countries: United States, Canada

• We are unable to support H1-B or STEM OPT candidates at this time.
• Located in the United States or Canada.

• Payments are processed weekly based on services rendered.